Assignment 3 - Data set Pathway and Network Analysis

Introduction

For ease of use, I wanted to include a link to my journal entry page that is complementary to this R Notebook here. My expression data set utilized in this report is GSE159559 which is a data set that may have therapeutic significance for treating adenocarcinoma, a form of cancer formed in the glands surrounding organs. Adenocarcinoma often metastasizes, dramatically decreasing patient’s likelihood of recovery. The experiment posits that FAM83H‐AS1, a noncoding driver of oncogenesis, might potentially be a therapeutic target for lung adenocarcinoma. The experiment provides 3 negative controls for A549 human cells, followed by down regulated A549 cells for FAM83H‐AS1, in hopes of showing that FAM83H-AS1 inhibits lung adenocarcinoma apoptosis.

Geo ID Used to Download Data: GSE159559

The Organism: Human

Number of GEO Data Sets That use the Same Technology: 327

Number of GEO Samples that use this Technology: 5113

Number of Genes Available Prior to Filtering: 40173

Standard for Filtering: Remove features without at least 1 read per million in the 3 replicates.

Number of Genes Available After Filtering: 15082

Method used for Normalization Trimmed Mean of M-values

Multiple Hypothesis Testing Method Benjamini-Hochberg method

P-Value Utilized 0.05

Thresholded Analysis Platform G-Profiler

Non-Thresholded Gene Set Enrichment Analysis

Using the ranked list of genes from assignment 2, I performed a non-thresholded gene set enrichment analysis, using GSEA (GSEA, in reference). I used the following inputs to run my GSE analysis.

#

Note that 1000 initial permutations are run from our ranked gene list, with a minimal size of 15 and maximal size of 200. I initially ran the GSEA using a default gene set from GSEA, but I decided to use the Bader gene set after having lackluster results with my network. Specifically, I used

HUMAN_GOBP_AllPathways_no_GO_iea_April_01_2022_symbol .gmt

(Bader, in reference.) Below is a generic resulting page from running the GSEA, for overview information from it.

#

It found that 3039 gene sets were upregulated, with 1280 gene sets showing significant FDR (<25%), and 1721 gene sets that were downregulated, with 524 genes showing significant FDR (<25%). The proportions between the non-thresholded GSEA analysis and the thresholded G-profiler analysis done in the previous assignment can be compared as follows (in a qualitative way). Their proportion of upregulated genes are in both analysis’ are significantly greater than the downregulated genes (from FAM83HS-AS1 silencing), but the most striking difference is that non-thresholded analysis contributed many more gene sets (and genes) than the thresholded one, which produced 1906 upregulated genes and 802 downregulated genes. This makes sense if you understand that the purpose of the non-thresholded gene analysis is to allow for many signals that came from genes that weren’t significantly over or under expressed to emerge. Of course, this comparison is not as straightforward as with GSEA as we are engaging directly with gene sets rather than directly with the manipulations we did with genes in G:Profiler.

Visualizing GSEA IN CYTOSCAPE

Immediately below, I insert the parameters used for making a non-thresholded gene set enrichment network, prior to any sort of manipulation:

#

The initial resulting map is shown here:

#

The resulting network has 904 nodes with 5032 edges, with the following thresholds (to make the network not overly populated.)

#

They were annotated using the defaults that EnrichmentMap applied to them, and were treated with a radial heatmap utilizing the following constraints, shown in the legend below for the publication ready figure:

#

Here is the publication ready figure:

#

Interpretation

Perhaps I should have raised the threshold for entry (solely on the network) because there is a multitude of competing themes present. Of note though, there are two massive themes present when analyzing. Massively upregulated connections pertaining to actin regulation and the change in regulation (usually upregulated) for nodes that pertain to the regulation of factors being produced for cell replication. As FAM83H‐AS1 is posited to be a noncoding oncogenic driver, which makes sense that pathways involving cell replication would be upregulated (Wang, in references). To add, FAM83H‐AS1is posited to be inhibitory towards apoptosis, so its silencing massively influencing the actin that produce the cytoskeleton is in line with our expectations. There is an article published in Nature (Campbell, in references). That suggests that actin cytoskeleton is a key regulator for apoptosis, which is in line with the original research statement that FAM83H-AS1 potentially inhibits cell apoptosis. Of course, this would need more thorough focused research to interogate. Overall, our research here is in line with the expectations the initial research article outlines; that FAM83H-AS1 has significant potential to be an therapeutic target for lung cancer.

Specific pathway

A gene of interest, RAB14, is discussed in the article (Wang, in references), as a potential therapeutic targets for lung adenocarcinoma, as it is an oncogene that were down-regulated in the cells silenced for FAM83H-AS1. The article never goes into depth as to why these are oncogenes of interest or what they might be involved with. I found in the network group pertaining to phagocytosis that RAB14 was collected in and well-connected to many other genes in the set. It is the most significantly down-regulated gene pertaining to phagocytosis, shown below, annotated using log-fold expression value changes:

#

Here we can see the plurality of interactions in the network come from physical interactions

#

The reason for investigating phagocytosis are two; RAB14 is the only gene experiencing significant downregulation in silenced FAM83H-AS1 cells, and more importantly, the processes of phagocytosis and apoptosis are related, albeit rather tenuously (L-P Erwig, in references.) Perhaps RAB14 has an influence on phagocytotic cells that would target the cancer-ridden cells, but that is just a thought and would really need more followup to examine.

References

Bader: Enrichment Map: A Network-Based Method for Gene-Set Enrichment Visualization and Interpretation Merico D, Isserlin R, Stueker O, Emili A, Bader GD PLoS One. 2010 Nov 15;5(11):e13984

Campell: Gourlay, C., Ayscough, K. The actin cytoskeleton: a key regulator of apoptosis and ageing?. Nat Rev Mol Cell Biol 6, 583–589 (2005). https://doi.org/10.1038/nrm1682

Cytoscape: Institute for Systems Biology, 2019. Cytoscape, Available at: https://www.cytoscape.org.

GSEA: Subramanian, Tamayo, et al. 2005 Proc Natl Acad Sci U S A 102(43):15545-50

L-P Erwig: Erwig, LP., Henson, P. Clearance of apoptotic cells by phagocytes. Cell Death Differ 15, 243–250 (2008). https://doi.org/10.1038/sj.cdd.4402184

Wang: Wang, S., Han, C., Liu, T., Ma, Z., Qiu, M., Wang, J., You, Q., Zheng, X., Xu, W., Xia, W., Xu, Y., Hu, J., Xu, L., & Yin, R. (2021). FAM83H-AS1 is a noncoding oncogenic driver and therapeutic target of lung adenocarcinoma. Clinical and translational medicine, 11(2), e316. https://doi.org/10.1002/ctm2.316